Improve Neural Distinguishers of SIMON and SPECK

 

Algorithm 1: Search for neural distinguisher based on SAT.

	Input: Network Architecture , Cipher with block size bits, Round .
	Output: Neural distinguisher , Input difference of distinguisher .
(1)	Search for the optimal probability as
(2)	Search for the differential characteristics with probability in , and save their input differences as
(3)	[ ]
(4)	[ ]
(5)	for in do
(6)	#Generate train set using d as the input difference
(7)	#Generate test set using d as the input difference
(8)	#Training using train data
(9)	#Get the accuracy of the model
(10)	ifthen
(11)
(12)
(13)	end if
(14)	end for
(15)	return